The photo is sourced from pronpz.ru
The purpose of electric centrifugal pumps is to bring oil from the well to the surface. In most cases, the monitoring of pump condition is carried out via complex telemetry systems encompassing dozens of gauges and enabling remote collection of information. However, these systems often fail in aggressive underground environments, requiring a complete shutdown of the oil production process.
The scientists from Perm Polytechnic University proposed to streamline the procedure for monitoring the underground operation of the pump by using only two sensors for the current and the voltage. This was made possible through computer modeling of the entire production process, which, in addition to an electric centrifugal pump, uses a control station, a transformer, cable lines and tubing. Voltage from the transformer is supplied via a cable line to the winding of the motor, which starts spinning, activating the electric centrifugal pump. In turn, the latter begins to pump oil from the well.
The computer model helped create a monitoring system that employs the so-called sigma-point Kalman filter, which uses differential equations to estimate the operating parameters of an oil production unit, such as currents, flux linkage, cable line resistance, load torque and shaft rotation speed of the submersible electric motor.
“The values from the current and voltage sensors are input to a program, which filters them, and we get a better signal as a result. This is how we can track all the parameters we need. A sigma-point Kalman filter is integrated into the control system, allowing us to control the submersible electric motor without using a sensor. We have now fully developed the software for the control system, and we plan to test it with real data later,” Rostislav Yudin, engineer at the Department of Microprocessor Automation, is quoted as saying by Perm Polytechnic University.
In addition to the computer model of the oil production unit, the authors of the study developed an oil yield monitoring system, which uses machine learning to predict the volume of oil recovered based on the motor shaft rotation speed, shaft load torque and oil density. This allowed the researchers to obtain a system for sensorless monitoring of the oil production process, which could reduce the cost of oil recovery.
